# megvii-research/NAFNet

The state-of-the-art image restoration model without nonlinear activation functions.

Repository: https://github.com/megvii-research/NAFNet
Canonical: https://ross.abutalabs.com/products/nafnet
Language: Python
License: NOASSERTION
License Family: other
Topics: deblur, denoise, stereo-super-resolution, eccv2022, image-deblurring, image-denoising, image-restoration, low-level-vision, pytorch
Last push: 2024-07-03T13:26:24+00:00

## Health v2 (maintenance only)
Score: 32/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 1606, "days_push": 791, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 3148, forks 400 (observed 2026-08-28T04:07:46.076158+00:00)

## What it is
NAFNet is the official PyTorch implementation of a state-of-the-art image restoration network that removes nonlinear activation functions. It provides pretrained models for image deblurring, denoising, and stereo image super-resolution.

## Use cases
- remove blur from photos
- denoise noisy images
- upscale stereo image pairs
- train a custom image restoration model
- benchmark restoration models on GoPro or SIDD
- restore low-quality images with a deep learning model

## When to choose
- you need state-of-the-art deblurring or denoising quality
- you want a simple, efficient restoration baseline in PyTorch
- you work on stereo image super-resolution

## When to avoid
- you need a general-purpose image editor rather than a research model
- you cannot run GPU inference
- you need a permissively licensed model (license is non-standard)

## Facets
- artifact type: library
- maturity: stable
- function: image-processing, machine-learning, deep-learning
- domain: computer-vision, image-processing, deep-learning, machine-learning
- platform: python
- tags: image-restoration, deblurring, denoising, super-resolution, pytorch, eccv2022, low-level-vision, gpu

## Member repositories
- megvii-research/NAFNet (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:46.076158+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T18:45:35.434279+00:00, confidence not recorded.
  - readme: https://github.com/megvii-research/NAFNet (fetched 2026-08-28T04:07:46.076158+00:00, sha ff22fd1bc832)
- Data as of 2026-08-30T08:39:29.467469+00:00.